Library › AI Evals › Managers and team leads
Set an accuracy bar per topic from the cost of a wrong answer
- For Managers and team leads
- You practice Catching mistakes and made-up facts
- Time about 5 minutes
- Level Beginner
About this exercise
Spokewise accepts at most $600 a month of wrong-answer cost per topic. For each topic, work out the cost if every answer were wrong, the share of answers that may be wrong ($600 divided by that cost), and the accuracy bar. Write exactly 4 numbered lines like "1. Hours: 6,000 x $2 = $12,000; $600 / $12,000 = 5%; bar 95%", then a last line "Strictest: [topic], [bar]". Plain text, under 110 words.
Common mistakes when you ask AI for this
- One bar for everything. The AI repeats the draft plan's single 95% bar. Try: Say: "Work out a separate bar per topic from its own cost."
- Skipping the working. The AI states 98% without showing how it got there. Try: Add: "Show the math on each line like 6,000 x $2 = $12,000."
- Mixing up the two steps. The AI divides the wrong way or treats the $600 as per chat. Try: Say: "Share allowed wrong = $600 / cost-if-all-wrong; bar = 100% minus that."
- No final verdict line. The AI ends with a paragraph of caveats. Try: Add: "Last line: Strictest: [topic], [bar]."
- A report instead of lines. The AI writes headings, bold and a long table. Try: Say: "Exactly 4 numbered lines then one last line, plain text."
What a strong prompt includes
- Asks for the math
- Uses the $600 budget
- Wants one bar per topic
- Gives a line format
- Asks for plain text
Related exercises
- Accrue unpaid wages for the working days after the last pay period (Finance AI · Accountants)
- Age a returns queue in business days and flag the ones over the limit (Customer Support & Success AI · Customer support teams)
- Alert sales to a competitor's price change, facts only (Research & AI Search · Product managers)
- Annualize recurring revenue without the one-time fee (Data AI · Finance and accounting)
- Answer a credit dispute with the full-hour rule and the math (Customer Support & Success AI · Customer support teams)